An intelligent fault diagnosis for centrifugal pumps based on electric current information available in industrial communication networks

Intelligent fault diagnosis systems of equipment are essential in industries, to prevent failures, reduce downtime and increase plant availability. In this context, this work proposes a methodology to develop diagnostic systems for detecting anomalies and identifying failures in the piping system due to malfunctioning pumps, such as cavitation. Relevant information is extracted based on motor electric current, available in PROFINET networks by smart relays, widely used in this type of application. A prototype of a hydraulic piping system was created for data collection and methodology validation. Machine learning tools for extracting and selecting features (such as standard deviation, entropy, harmonics, among others), and an SVM-based model for pattern recognition, are used to diagnose the status of the centrifugal pump. The results achieved were 88.7 % of accuracy for cavitation fault detection and 100.0% for dry running operation. Thus, the methodology is feasible and efficient, indicating an opportunity to improve the use of data available by smart relays in industrial network communication.